User-level communication in cluster-based servers

E. V. Carrera, S. Rao, L. Iftode, R. Bianchini

Research output: Chapter in Book/Report/Conference proceedingConference contribution

25 Citations (Scopus)

Abstract

Clusters of commodity computers are currently being used to provide the scalability required by several popular Internet services. In this paper we evaluate an efficient cluster-based WWW server, as a function of the characteristics of the intra-cluster communication architecture. More specifically, we evaluate the impact of processor overhead, network bandwidth, remote memory writes, and zero-copy data transfers on the performance of our server. Our experimental results with an 8-node cluster and four real WWW traces show that network bandwidth affects the performance of our server by only 6%. In contrast, user-level communication can improve performance by as much as 29%. Low processor overhead, remote memory writes, and zero-copy all make small contributions towards this overall gain. To be able to extrapolate from our experimental results, we use an analytical model to assess the performance of our server under different workload characteristics, different numbers of cluster nodes, and higher performance systems. Our modeling results show that higher gains (of up to 55%) can be accrued for workloads with large working sets and next-generation servers running on large clusters.

Original languageEnglish (US)
Title of host publicationProceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002
PublisherIEEE Computer Society
Pages275-286
Number of pages12
ISBN (Electronic)0769515258
DOIs
StatePublished - Jan 1 2002
Event8th International Symposium on High-Performance Computer Architecture, HPCA 2002 - Cambridge, United States
Duration: Feb 2 2002Feb 6 2002

Publication series

NameProceedings - International Symposium on High-Performance Computer Architecture
Volume2002-January
ISSN (Print)1530-0897

Other

Other8th International Symposium on High-Performance Computer Architecture, HPCA 2002
CountryUnited States
CityCambridge
Period2/2/022/6/02

Fingerprint

Servers
Communication
World Wide Web
Bandwidth
Data storage equipment
Data transfer
Scalability
Analytical models
Internet

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture

Keywords

  • Analytical models
  • Computer architecture
  • Computer science
  • Network servers
  • Read-write memory
  • Scalability
  • Stress
  • Web and internet services
  • Web server
  • World Wide Web

Cite this

Carrera, E. V., Rao, S., Iftode, L., & Bianchini, R. (2002). User-level communication in cluster-based servers. In Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002 (pp. 275-286). [995717] (Proceedings - International Symposium on High-Performance Computer Architecture; Vol. 2002-January). IEEE Computer Society. https://doi.org/10.1109/HPCA.2002.995717
Carrera, E. V. ; Rao, S. ; Iftode, L. ; Bianchini, R. / User-level communication in cluster-based servers. Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002. IEEE Computer Society, 2002. pp. 275-286 (Proceedings - International Symposium on High-Performance Computer Architecture).
@inproceedings{713649bcacef45c4b6e904ba239c45ba,
title = "User-level communication in cluster-based servers",
abstract = "Clusters of commodity computers are currently being used to provide the scalability required by several popular Internet services. In this paper we evaluate an efficient cluster-based WWW server, as a function of the characteristics of the intra-cluster communication architecture. More specifically, we evaluate the impact of processor overhead, network bandwidth, remote memory writes, and zero-copy data transfers on the performance of our server. Our experimental results with an 8-node cluster and four real WWW traces show that network bandwidth affects the performance of our server by only 6{\%}. In contrast, user-level communication can improve performance by as much as 29{\%}. Low processor overhead, remote memory writes, and zero-copy all make small contributions towards this overall gain. To be able to extrapolate from our experimental results, we use an analytical model to assess the performance of our server under different workload characteristics, different numbers of cluster nodes, and higher performance systems. Our modeling results show that higher gains (of up to 55{\%}) can be accrued for workloads with large working sets and next-generation servers running on large clusters.",
keywords = "Analytical models, Computer architecture, Computer science, Network servers, Read-write memory, Scalability, Stress, Web and internet services, Web server, World Wide Web",
author = "Carrera, {E. V.} and S. Rao and L. Iftode and R. Bianchini",
year = "2002",
month = "1",
day = "1",
doi = "10.1109/HPCA.2002.995717",
language = "English (US)",
series = "Proceedings - International Symposium on High-Performance Computer Architecture",
publisher = "IEEE Computer Society",
pages = "275--286",
booktitle = "Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002",
address = "United States",

}

Carrera, EV, Rao, S, Iftode, L & Bianchini, R 2002, User-level communication in cluster-based servers. in Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002., 995717, Proceedings - International Symposium on High-Performance Computer Architecture, vol. 2002-January, IEEE Computer Society, pp. 275-286, 8th International Symposium on High-Performance Computer Architecture, HPCA 2002, Cambridge, United States, 2/2/02. https://doi.org/10.1109/HPCA.2002.995717

User-level communication in cluster-based servers. / Carrera, E. V.; Rao, S.; Iftode, L.; Bianchini, R.

Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002. IEEE Computer Society, 2002. p. 275-286 995717 (Proceedings - International Symposium on High-Performance Computer Architecture; Vol. 2002-January).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - User-level communication in cluster-based servers

AU - Carrera, E. V.

AU - Rao, S.

AU - Iftode, L.

AU - Bianchini, R.

PY - 2002/1/1

Y1 - 2002/1/1

N2 - Clusters of commodity computers are currently being used to provide the scalability required by several popular Internet services. In this paper we evaluate an efficient cluster-based WWW server, as a function of the characteristics of the intra-cluster communication architecture. More specifically, we evaluate the impact of processor overhead, network bandwidth, remote memory writes, and zero-copy data transfers on the performance of our server. Our experimental results with an 8-node cluster and four real WWW traces show that network bandwidth affects the performance of our server by only 6%. In contrast, user-level communication can improve performance by as much as 29%. Low processor overhead, remote memory writes, and zero-copy all make small contributions towards this overall gain. To be able to extrapolate from our experimental results, we use an analytical model to assess the performance of our server under different workload characteristics, different numbers of cluster nodes, and higher performance systems. Our modeling results show that higher gains (of up to 55%) can be accrued for workloads with large working sets and next-generation servers running on large clusters.

AB - Clusters of commodity computers are currently being used to provide the scalability required by several popular Internet services. In this paper we evaluate an efficient cluster-based WWW server, as a function of the characteristics of the intra-cluster communication architecture. More specifically, we evaluate the impact of processor overhead, network bandwidth, remote memory writes, and zero-copy data transfers on the performance of our server. Our experimental results with an 8-node cluster and four real WWW traces show that network bandwidth affects the performance of our server by only 6%. In contrast, user-level communication can improve performance by as much as 29%. Low processor overhead, remote memory writes, and zero-copy all make small contributions towards this overall gain. To be able to extrapolate from our experimental results, we use an analytical model to assess the performance of our server under different workload characteristics, different numbers of cluster nodes, and higher performance systems. Our modeling results show that higher gains (of up to 55%) can be accrued for workloads with large working sets and next-generation servers running on large clusters.

KW - Analytical models

KW - Computer architecture

KW - Computer science

KW - Network servers

KW - Read-write memory

KW - Scalability

KW - Stress

KW - Web and internet services

KW - Web server

KW - World Wide Web

UR - http://www.scopus.com/inward/record.url?scp=84949773027&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84949773027&partnerID=8YFLogxK

U2 - 10.1109/HPCA.2002.995717

DO - 10.1109/HPCA.2002.995717

M3 - Conference contribution

AN - SCOPUS:84949773027

T3 - Proceedings - International Symposium on High-Performance Computer Architecture

SP - 275

EP - 286

BT - Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002

PB - IEEE Computer Society

ER -

Carrera EV, Rao S, Iftode L, Bianchini R. User-level communication in cluster-based servers. In Proceedings - 8th International Symposium on High-Performance Computer Architecture, HPCA 2002. IEEE Computer Society. 2002. p. 275-286. 995717. (Proceedings - International Symposium on High-Performance Computer Architecture). https://doi.org/10.1109/HPCA.2002.995717